NATURE|Vol 455|23 October 2008 NEWS FEATURE

DISPUTED DEFINITIONS
If you want to start an argument, ask the person who just said ‘paradigm shift’ what it really
means. Or ‘epigenetic’. Nature goes in search of the terms that get scientists most worked up.
o a great extent, science is about arriving at defini- comes along and overturns the previous consensus. Voilà,

ILLUSTRATIONS: N. DEWAR
tions. What is a man? What is a number? Questions a paradigm shift. The classic example, Kuhn said, is the
such as these require substantial inquiry. But where Copernican revolution, in which Ptolemaic theory was swept
science is supposed to be precise and measured, away by putting the Sun at the centre of the Solar System.
definitions can be frustratingly vague and variable. Post-shift, all previous observations had to be reinterpreted.
Here, Nature looks at some of the most difficult Kuhn’s theory about how science works was arguably a
definitions in science. Some are stipulative definitions, cre- paradigm shift of its own, by changing the way that academ-
ated by scientists for their convenience, but on which the ics think about science. And scientists have been using the
community has not found consensus. Popular though they phrase ever since.
are, not everyone agrees on what is meant by ‘paradigm In a postscript to the second edition of his book, Kuhn
shift’ or ‘tipping point’. explained that he used the word ‘paradigm’ in at least two
Essential definitions — those that get at the ways (noting that one “sympathetic reader” had found 22
question of what makes a thing a thing — can uses of the term). In its broad form, it encompasses the
be just as troublesome. What is race, or con- “entire constellation of beliefs, values, techniques and so
sciousness? And does it even matter if there on shared by the members of a given community”. More
is no agreed-on meaning? specifically it refers to “the concrete puzzle-solutions” that
The good news is that for every troublesome are used as models for normal science post-shift.
term there are thousands used every day with Scientists who use the term today don’t usually mean
no problems. Scientists are competent, if uncon- that their field has undergone a Copernican-scale revolu-
scious wielders of definition, says Anil Gupta, a tion, to the undying annoyance of many who hew to Kuhn’s
philosopher of science at the University of Pitts- narrower definition. But their usage might qualify under his
burgh in Pennsylvania, “just as one can walk quite broader one. And so usage becomes a matter of opinion
happily without having a complete account of walking”. and, perhaps, vanity.
The use of the term in titles and abstracts of
leading journals jumped from 30 papers in 1991 to 124
Paradigm shift in 1998, yet very few of these papers garnered more
[ ] noun. than 10 citations apiece1. Several scientists contacted for
this article who had used paradigm shift said that, in
aradigm shift has a definite origin and originator: retrospect, they were having second thoughts. In
P Thomas Kuhn, writing in his 1962 book The Structure
of Scientific Revolutions, argued against the then prevalent
2002, Stuart Calderwood, an oncologist at Harvard
Medical School in Boston, Massachusetts, used it
view of science as an incremental endeavour marching ever to describe the discovery that ‘heat shock proteins’,
truthwards. Instead, said Kuhn, most science is “normal crucial to cell survival, could work outside the cell
science”, which fills in the details of a generally accepted, as well as in2. “If you work in a field for a long time
shared conceptual framework. Troublesome anoma- and everything changes, it does seem like a revolu-
lies build up, however, and eventually some new science tion,” he says. But now he says he may have misused
1023
NEWS FEATURE NATURE|Vol 455|23 October 2008

the phrase because the discovery was adding to, rather than

NOAA GOES-7/NASA
overturning, previous knowledge in the field.
Arvid Carlsson, of the University of Gothenburg in
Sweden stands by his use of the phrase. “Until a certain time,
the paradigm was that cells communicate almost entirely
by electrical signals,” says Carlsson. “In the 1960s and ’70s,
this changed. They do so predominantly by chemical sig-
nals. In my opinion, this is dramatic enough to deserve the Ptashne of Memorial Sloan–Kettering Cancer Center in
term paradigm shift.” Few would disagree: base assump- New York, who has published his own take on the word’s
tions were overturned in this case, and Carlsson’s own work usage4. “I’ve grown to be very careful about using the term,”
on the chemical neurotransmitter dopamine (which was says Bing Ren, who studies gene regulation at the University
instrumental in this particular shift) earned him the 2000 of California, San Diego.
Nobel Prize in Physiology or Medicine According to the ‘traditional’ definition that Ptashne
Unless a Nobel prize is in the offing, it might be wise for favours, epigenetics describes “a change in the state of expres-
scientists to adopt the caution of contemporary historians of sion of a gene that does not involve a mutation, but that is
science and think twice before using a phrase with a complex nevertheless inherited in the absence of the signal or event
meaning and a whiff of self promotion. “Scientists all want to that initiated the change”. The classic example is found in a
be the scientists that generate a new revolution,” says Kuhn’s bacteriophage called Lambda, which stays dormant in the
biographer, Alexander Bird, a philosopher at the University genome of generations of cells through inheritance of a regu-
of Bristol, UK. “But if Kuhn is right, most science is normal latory protein. These sort of processes are basic to some of the
science and most people can’t perform that role.” most pressing questions in biology today: such mechanisms
Emma Marris
“Epigenetics is are needed to explain how a single-celled embryo can gener-
a useful word if ate cells that are genetically identical, but express a different
you don’t know array of genes and hence take on different jobs in blood, brain
Epigenetic what’s going on or muscle for generation after generation.
[ ] adjective. Over the past few years, however, all kinds of processes
— if you do, you associated with gene control have been subsumed under
o one denies that epigenetics is fashionable: its usage use something the moniker. For example, ‘epigenetic’ is often used to refer
N in PubMed papers increased by more than tenfold
between 1997 and 2007. And few deny that epigenetics is
else.” to the chemical modification of histones — proteins that
DNA winds around — which is involved in gene regulation.
important. What they do disagree on is what it is. — Adrian Bird This infuriates those who learned the classical definition;
“The idea is that there is a clear meaning and that it’s being they say it puts these modifications at the heart of devel-
violated by people like me who use it more loosely,” says opment and disease despite scant evidence that they are
Adrian Bird at the University of Edinburgh, UK . Last year inherited. “Why did histone marks become epigenetic?”
he suggested this as a definition: “the structural adaptation says Kevin Struhl at Harvard Medical School in Boston.
of chromosomal regions so as to register, signal or perpetu- “People decided that if they call them that it makes them
ate altered activity states”3. But this wide-ranging proposal, interesting.” Others say that it is not about making things
which takes on-board pretty much every physical indica- sound important, it is more the lack of any other phrase
tor of a gene’s activity is “preposterously dumb”, says Mark with which to collectively refer to this type of work.
The word had dual meanings long before the current
debate. In the 1940s, Conrad Waddington used it to describe
how the genetic information in a ‘genotype’ manifests itself as
a set of characteristics, or ‘phenotype’. In 1958, David Nanney
at the University of Michigan, Ann Arbor, borrowed the term
to describe “messy” inherited phenomena that could not be
explained by conventional genetics5. “It was controversial in
1958 and everything died down and it has come alive again,”
says Nanney. “It took 40 years for epigenetics to become a
major word in the vocabulary of modern biology.”
A lot of money can ride on whether a researcher is, or is not,
studying epigenetics: the US National Institutes of Health
(NIH) this month started handing out US$190 million
as part of its epigenomics initiative and other countries are
pouring funding into the area. (The NIH is careful to define
the epigenetics it is paying for as including both heritable
and non-heritable changes in gene activity, something
that Ptashne describes as “a complete joke”.) Bird says he
remains in favour of a relaxed usage. “Epigenetics is a useful
word if you don’t know what’s going on — if you do, you use
something else,” he says.
Helen Pearson
1024
 NATURE|Vol 455|23 October 2008 NEWS FEATURE

species, vehicles, neurons) that are constantly interacting

Complexity with, and adapting to, one another. They all display a rich
[ ] noun.
array of nonlinear feedback loops among the agents, which
means that small changes can have a big effect. And they
n his book Programming the Universe, never quite settle down into static equilibrium.
I engineer Seth Lloyd of the Massachusetts
Institute of Technology in Cambridge describes
The effort to understand complex systems has led
researchers to develop new analytical tools such as
how he once compiled 42 definitions of com- network theory, agent-based modelling and genetic
plexity — none of which encompasses every- algorithms. These tools, combined with the expo-
thing people mean by that word. Researchers nential growth in computational power, have allowed
in the many institutes and programmes researchers to build ever more complex models of
formed to study ‘complexity’ are still complex systems — and study the subtle but powerful
searching for the right way to describe phenomenon of ‘emergence,’ in which multiple agents
their discipline. “If we’re a univer- exhibit collective behaviour that is a great deal more
sity centre, we should be able to than the sum of its parts.
say what we care about,” says Carl So even though the field seems little closer to defin-
Simon, director of the Center for ing its subject, says Lloyd, “in places where people
the Study of Complex Systems at can apply these conceptual and computational tools,
the University of Michigan. we’ve made huge progress in understanding complex
The quest for a rigorous definition systems”. But in a world where we are constructing
reached a particularly intense pitch in ever more complex artefacts — technolo-
the early 1990s, when some of the more gies, economies, organizations and soci-
visionary researchers at the Santa Fe Insti- eties — even better tools are needed to
tute in New Mexico held out the hope of a universal theory keep pace.
of complexity — a mathematically precise set of equations M. Mitchell Waldrop
that would hold for all complex systems in much the same
way that the second law of thermodynamics holds for all
physical systems. Race
James Crutchfield, head of the Complexity Sciences [ ] noun.
Center at the University of California, Davis, says that this
created a problem. “New people would come into the field f biologists had a list of four-
and start using the word ‘complexity’ as if it was a unitary
thing” — which, as became increasingly clear, it was not. No
I letter words to avoid, then
‘race’ would be higher up than
all-encompassing theory emerged. Even within the precise anything more convention-
world of binary code and bit strings, there was computa- ally vulgar. It is controversial,
tional complexity, which describes how much memory and it lacks a clear definition and
processing is required to carry out a calculation; algorithmic the more that genetics reveals
complexity, which is related to how much a digital descrip- about race, the more biologi-
tion of something can be compressed; and any number of cally meaningless the term seems.
combinations and variations. “So my bottom line is, add an Race was long used to imply a shared, distinct ances-
adjective to ‘complexity’,” Crutchfield says. try, as in a 1936 definition of the term in Nature: “It
Researchers have found plenty of undeniably complex has two main connotations, one being community of
systems to study, such as economies, ecosystems, urban descent, the other distinctness from other races.” But in
traffic and brains (see ‘Consciousness’). And in a qualitative 1972, Harvard geneticist Richard Lewontin showed that
sense, at least, these systems do have certain features in com- the concept of race starts to dissolve under genetic scrutiny.
mon that might serve as a definition. They are, for instance, He found that the vast majority of human genetic varia-
all composed of many independent ‘agents’ (consumers, tion, which he measured in 16 genes, is found within, not
between, what he called the ‘classical racial groupings’6.
Since then, studies examining hundreds or even thousands
of genetic markers have confirmed Lewontin’s findings7,8.
A consensus now exists across the social and biological
sciences: regardless of appearance or heritage, groups of
human beings are overwhelmingly more genetically similar
G. HOLLAND/PHOTOLIBRARY.COM

to each other than different. This doesn’t mean race does

not exist or is meaningless in society — far from it. But it
does mean that an individual’s race is not a particularly
useful or predictive indicator of biological traits or medical
vulnerabilities. Race is “the social interpretation of how
we look, in a race-conscious society”, says Camara Phyllis
Jones, the research director on the Social Determinants of
Health and Equity programme at the US Centers for Disease
1025
NEWS FEATURE NATURE|Vol 455|23 October 2008

Control and Prevention in Atlanta, Georgia. Lewontin

says that assigned races are essentially arbitrary. “It means
essentially a group of related people, and where you draw
the line depends on where you are in history.”
Some argue that severing biology from the definition of
race risks jettisoning medically meaningful information.

BLUESTONE/SPL
Patterns of genetic variation can be used to classify people
from different geographical regions into clusters that some-
times mimic the classical racial groupings, and geneticists
say that members of these groups seem to have distinctive
disease prevalences and drug metabolism. So race could
serve as a cheap, albeit imperfect, surrogate for defining
groups for clinical trials or medical interventions.
But genetics is turning up ever more examples of how
race obscures relevant information. A study published an infectious disease’s ‘reproductive rate’ goes above one.
in April showed that a mutation found in 40% of African This means that each infected person infects more than one
Americans acts like an endogenous beta blocker to protect other and the disease starts growing into an epidemic.
patients with heart failure from death9. It also suggested The phrase reached its own tipping point in 2000 when
why previous research had found conflicting evidence Malcolm Gladwell, a staff writer at The New Yorker, pub-
about the response of African Americans to beta block- lished his successful book The Tipping Point: How Little
ers: those studies had lumped all African Americans into Things Can Make a Big Difference. It also acquired a wor-
one group, obscuring the effects of mutations that confer risome — some say alarmist — flavour courtesy of its fre-
protection or vulnerability. quent usage in the context of climate change.
A person’s perception of his or her race can still serve to Regarding climate, the term is commonly defined as the
capture life experiences relevant to behavioural and clinical critical threshold at which a slow gradual change qualita-
research, such as the stress of lifelong discrimination that tively alters the state of an entire system. This is different
may contribute to health disparities. But in other contexts to a ‘point of no return’ which is, by definition, irreversible.
researchers are abandoning the term in favour of other Only if internal forcing will cause a runaway effect is a tip-
ways to group humans, by ‘population,’ genetic ancestry’ or ping point also a point of no return.
‘geographic ancestry’. The idea that positive feedbacks — such as the melt-
Erika Check Hayden ing of polar ice reducing surface reflectivity, thereby
causing yet more solar absorption, warming and
melting — could amplify climate change to a point
Tipping point of fundamentally altering the global system has
[ ] noun. been around for decades. The debate now is about
where those tipping points lie, and what will hap-
n July 2006, scientists running the pen when they are crossed.
I RealClimate blog ironically head-
lined one of their posts ‘Runaway tip-
In a paper published in February, a team led by
Lenton looked at 15 potential tipping ‘elements’
ping points of no return’. The post laments (things that could reach tipping points) in Earth’s
that usage of the phrase ‘tipping point’ in climate system10. Arctic sea-ice and the Greenland
climate-change and ecosystem discussions ice sheet were those most at risk from ‘tipping’ within the
had reached, well — a tipping point. twenty-first century, the authors concluded.
It’s not the frequency of the word that But researchers accept that most known tipping points
bothers researchers. It’s the lack of one clear defi- seem to be reversible on human timescales. Melting of
nition and the confusing way in which the concept the complete Arctic summer ice sheet, for example, could
is being used, among scientists and in the public probably be reversed within a few years or so in a cooler
discourse, often to imply that global warming- world. Melting of the extremely thick Greenland and
induced changes will propel Earth into irreversible Antarctic ice sheets are a possible exception because, once
and catastrophic climate change. “There is no con- melted, new ice would have to form at lower, warmer
vincing theoretical argument or model that at some altitudes with less snowfall.
point the planet as a whole will snap into a second state Claims that global warming could reach an
of system,” says Timothy Lenton, an Earth scientist at the irreversible tipping point by 2016, as made last year by
University of East Anglia, UK. James Hansen, director of NASA’s Goddard Institute
The term was originally coined in 1958 by sociologist for Space Studies in New York, refer to the trajectory
Morton Grodzins in the context of studies on the racial of greenhouse-gas emissions, not to changes in the
makeup of US neighbourhoods. He found that when the climate system. Even if greenhouse-gas concentra-
migration of African-Americans into traditionally white tions reach a point at which they cannot be restored
neighbourhoods had reached a certain level, whites began to pre-industrial levels, it will not inevitably push the
to move out. In the 1970s, epidemiologists adopted tipping world’s climate over a catastrophic tipping point.
point to describe the threshold at which, mathematically, Quirin Schiermeier
1026
NATURE|Vol 455|23 October 2008 NEWS FEATURE

F. SAURER/SPL
Graf of the Centre for Genomic Regulation in Barcelona.
Some researchers are side-stepping the debate by refer-
ring in their papers to ‘stem/progenitor cells’. Fully under-
standing what each cell can do is more important than
knowing what to call them all, says Goldstein. “Some of
this just breaks down,” he says. “That’s biology. It wasn’t
designed to fit the language.”
Monya Baker

Significant
Stem cell [ ] adjective.
[ ] noun.
ew words in the scientific lexicon are as confusing, or as

A sk a group of stem-cell biologists to define stem cell,

and they’ll say roughly the same thing: a cell that can,
F loaded, as ‘significant’. Statisticians wring their hands over
its cavalier use to describe scientific validity. And backed by
long term, divide to make more copies of itself as well as statistics or not, researchers commonly employ the word
cells with more specialized identities. Ask the same scien- to illustrate the importance of their latest finding.
tists to list the most disputed terms in the field, however, The very definition of statistical significance is misunder-
and ‘stem cell’ will be top of that list. stood by most scientists, says Steven Goodman, a biostatisti-
The problem here is an operational one: reasonable people cian at the Johns Hopkins School of Medicine in Baltimore,
disagree on which cells qualify under the definition. “It’s not “Some of this Maryland, and associate editor on Annals of Internal Medi-
unusual to pick up a paper and see someone call something just breaks cine. Typically, researchers take a result to be statistically
a stem cell and the evidence that it is, is just not there,” says significant based on ‘p-values’. This parameter is used, for
Lawrence Goldstein, who directs the stem-cell research down. That’s example, to reveal whether a drug lowers cholesterol based
programme at the University of California, San Diego. biology. It wasn’t on promising data collected in a clinical trial.
Alleged ‘stem cells’ can fail to meet the definition on many designed to fit According to the common interpretation, a ‘significant’
counts. Stem cells should persist long term, yet many ‘stem result with a p-value of 0.05 or less means that there is a 5%
cells’ exist only in the fetus. Multipotency — the ability to the language.” or less chance that the drug is ineffective. According to the
generate multiple cell types — is a criterion for a haemat- — Lawrence statistically accurate definition, there is a 5% or less chance
opoietic, or blood-forming, stem cell, but spermatagonial Goldstein of seeing the observed data even though the drug is, indeed,
stem cells only produce sperm. Stem cells specific to tis- ineffective. Rhetorically, the difference may seem imper-
sue such as cartilage, the kidney and the cornea have been ceptible; mathematically, say statisticians, it is crucial. In
reported, with varying degrees of acceptance. The quest for situations in which the data is somewhat ambiguous, there
a ‘stemness signature’, a collection of markers common to is a chance that results can be misinterpreted. “It’s diaboli-
all stem cells, has been met with frustration. cally tricky,” Goodman says.
Debate erupts most commonly over whether a Most statisticians resign themselves to abuse of the term’s
particular cell should be considered a stem cell, which can strict definition. But more grievous trespasses abound.
divide indefinitely, or a progenitor cell, which can differen- “Statistical significance is neither a necessary nor a suffi-
tiate into fewer cell types and is thought to burn itself out cient condition for proving a scientific result,” says Stephen
after a certain number of divisions. Ziliak, an economist at Roosevelt University in Chicago,
The only way to be really sure of what a cell can, and can- Illinois, and co-author of The Cult of Statistical Signifi-
not do, is to observe it, but it is difficult to study cells in vivo, cance. P-values are often used to emphasize the certainty
and putting them in a dish might change their behaviour.
Haematopoietic stem cells were the first to be identified
and have, to some extent, set default standards. Putative
stem cells are isolated, then placed into animals whose own
haematopoietic stem cells have been destroyed by radiation.
If the blood-forming system is restored, the transplant is
assumed to have contained stem cells. But such an assay
is impossible when working with other cell types, such as
neural stem cells, which are harder to transplant and assess
in disease models. And it is difficult to pin the label to one
cell type, when studies commonly involve a mixed popula-
tion. “It is perhaps not realistic to come up with a generally
applicable definition of an adult stem cell,” says Thomas
1027
NEWS FEATURE NATURE|Vol 455|23 October 2008

switch off life support, but clinicians readily acknowledge

that the tests break down when patients are unable to move.
Doctors now find themselves in an uncomfortable limbo,
because it is not clear whether cortical activity measured on
fMRI is enough to redefine those decision points. “What do
of data, but they are only a passive read-out of a statistical we do as a community as long as this method is not yet vali-
test and do not take into account how well an experiment dated?” asks Steven Laureys, a neurologist with the Coma
was designed. A p-value would not reveal, for example, that Science Group at the University of Liége in Belgium.
everyone was taking different doses of that cholesterol drug. The French philosopher René Descartes declared
In many experiments, Ziliak says, “there are so many differ- that consciousness was a fundamental property that fell
ent errors that they tend to swamp the p-value errors”. beyond the rules of the physical world. Most scientists, says
Even if a result is a genuinely statistically significant one, Edelman, are not satisfied with that answer. “There must
it can be virtually meaningless in the real world. A new be some physical basis for consciousness,” he says. “The
cancer treatment may ‘significantly’ extend life by a month, difficulty is, how does that arise?”
but many terminally ill patients would not consider that Philosophers David Chalmers of the Australian National
outcome significant. A scientific finding may be ‘significant’ University in Canberra, explored what he called the “hard
without having any major impact on a field; conversely, problem” of consciousness by pondering‘qualia’, the subjec-
the significance of a discovery might not become apparent tive properties of experiences. Scientists and philosophers
until years after it is made. “One has to reserve for history alike have struggled to explain how the physical properties
the judgement of whether something is significant with “You don’t of the world around us — such as colour and temperature —
a capital S,” says Steven Block, a biophysicist at Stanford give rise to the experiences of ‘red’, or ‘warm’. Chalmers has
University in California. waste your time argued that the functional organization of the brain rather
In some situations other statistical methods can substitute, defining the than its chemical or molecular properties makes these expe-
but Goodman believes that trying to use them in the scien- thing. You just riences possible.
tific literature would be like “talking Swahili in Louisiana”. Many definitions of consciousness include the ability
He says he and other editors do their best to keep the term go out there and to sort through the relentless onslaught of incoming data
out of Annals though. “We ask them to use words like ‘statis- study it.” to create and respond to an internal model of the external
tically detectable’ or ‘statistically discernable,’” he says. — Michael world. And some believe that simply gathering data about
Geoff Brumfiel neurons and behaviours will not be enough. “What we need
Gazzaniga is a ‘theory of consciousness’. Then we’ll be in a better posi-
tion to define it,” says professor of biology and engineering
Consciousness Christof Koch of the California Institute of Technology in
[ ] noun. Pasadena. Koch thinks that information theory could pro-
vide the solution by determining whether consciousness
sychologists, philosophers, neurobiologists and doctors might be an inherent by-product of a system as enormously
P all grapple with the term consciousness. For clinicians,
the definition is of life or death importance; for some oth-
complex as the brain (see ‘Complexity’).
Michael Gazzaniga, a neuroscientist at the
ers, it is a matter of determining how the brain’s intercon- University of California, Santa Barbara, argues
necting tissues collectively create a sense of self. “How can that researchers need only develop a work-
this three-pound piece of meat inside my head give rise ing definition to explore consciousness, not
to something like being me?” sums up Gerald Edelman, a precise one. “You don’t waste your time
director of the Neurosciences Insti- defining the thing,” he says. “You just go
tute in La Jolla, California. out there and study it.” ■
In 2006, neuroscientist Adrian Heidi Ledford
Owen, at the Medical Research
Council Cognition and Brain 1. Cohen, J. Science 283, 1998–1999 (1999).
2. Asea, A. et al. J. Biol. Chem. 277, 15028–15034
Sciences Unit in Cambridge, UK, (2002).
reported that a woman who had been 3. Bird, A. Nature 447, 396–398 (2007).
diagnosed as being in a vegetative 4. Ptashne, M. Curr. Biol. 17, R233–R236 (2007).
5. Nanney, D. L. Proc. Natl Acad. Sci. USA 44,
state had shown signs of brain activity 712–717 (1958).
associated with consciousness11. The 6. Lewontin, R. C. Evol. Biol. 6, 381–398 (1972).
activity was picked up with functional 7. Li, J. Z. et al. Science 319, 1100–1104 (2008).
8. Jakobsson, M. et al. Nature 451, 998–1003
magnetic resonance imaging (fMRI), (2008).
which can reveal changes in brain 9. Liggett, S. B. et al. Nature Med. 14, 510–517
blood flow. (2008).
The finding rattled the clinical defini- 10. Lenton, T. M. et al. Proc. Natl Acad. Sci. USA 105,
1786–1793 (2008).
tion of consciousness, which is determined 11. Owen, A. M. et al. Science 313, 1402 (2006).
by using a series of behavioural tests to see if the
patient can make voluntary movements in response See Essay, page 1040.
to commands. The outcome can determine whether Discuss definitions online at http://
a patient needs pain medication, or whether it is time to tinyurl.com/4afapl.

1028